Psychometric properties of the final forms of the ARIK-A
Alternate-form reliability
For a subset of the examinees who completed both forms, we selected either form A or B in an
alternating fashion. Data indicate that the yoked items from Forms A and B likely yield similar
results. A correlation coefficient of 0.84 was obtained between the Forms A and B composites
(Reading Instructional Knowledge Composite or RIKC) (n0120). All alternate forms coefficients
are in Table 2; scale reliabilities range between 0.71 (Fluency scale) to 0.79
(Alphabetics). Alternate form reliability coefficients for the two global scales are: Basic, 0.79
and Advanced, 0.82. In addition, there are no statistically significant differences between the
Form A and Form B means on any of the scales or composite scores (p>0.05). Additional
evidence that items function similarly on both forms is available from examination of the
difficulty levels (p values) for the yoked items from Forms A and B. In fact, difficulty levels for
the yoked items from Forms A and B are virtually identical, and the average p values are 0.62
and 0.61 for the two forms, respectively. The modal p value difference of yoked items is 0.01.
The strong similarity of the p values for the yoked items provides evidence for use of the two
forms in a pretest–posttest format (see the manual for actual p values).